Professional Documents
Culture Documents
Asmae El Mokrini
Rabat Business School, BEAR-Lab,
International University of Rabat, Morocco
Email: asmae.elmokrini@uir.ac.ma
Youssef Boulaksil
College of Business and Economics,
UAE University, Al Ain, United Arab Emirates
Email: y.boulaksil@uaeu.ac.ae ; youssef.boulaksil@gmail.com (Corresponding Author)
Abdelaziz Berrado
Research team AMIPS, EMI,
Mohammed V University, Rabat, Morocco
Email: berrado@emi.ac.ma
problems (Melo et al., 2009, Eiselt and Marianov, 2011, reform can produce different results in different countries
Arabani and Farahani, 2012, Drezner and Hamacher, 2001, (Roberts, 2011). Arney et al. (2014) state that the
Snyder, 2006), Şahin, and Süral, 2007). Facility location pharmaceutical distribution system in particular in most
models can be generally classified into these four categories: emerging markets has a different market structure compared
analytical models, continuous models, network models, and to developed countries. The main differences include a lack
discrete models (ReVelle et al. 2008). Analytical of distribution networks with national reach, excessive
modelsdescribe systems as mathematical models by making fragmentation and many small players, many intermediaries
simplified assumptions regarding the demand distribution or between the manufacturer and the patient, and poor IT and
the cost structure of the network while giving exact solutions communication flow systems resulting in poor coordination
(Mousavi et al, 2013). In continuous models, facilities can across actors in the distribution channel. Facility location
be located at any point on a continuous space while demand modelling can then be affected by the differences between
is typically concentrated in points (Carlsson and Jia, 2014). emerging markets and developed countries regarding the
The selection of sites for new facilities in discrete models is nature of facilities and their sector of activity.
limited to a finite set of available candidate locations (Nickel The literature shows several implementations of facility
and Gama, 2015; Kochetov, 2011; Schütz et al., 2008). location models pertaining to the healthcare sector in
Network problems, on the other hand, present models where developed countries (Beheshtifar and Alimoahmmadi, 2015;
facilities and customers are positioned on a network and Laporte et al., 2015; Haase and Muller, 2015; Han, 2014;
demand is typically associated to nodes (Daskin, 2011). In Khodaparasti et al., 2016). Drezner and Hamacher (2001)
approaching facility location problems in the specific context state that the implementation of location-allocation problems
of healthcare (e.g. Mousazadeh et al., 2018), the literature in emerging markets is different from that of developed
shows that the three most used models are the location set countries, mainly because of data limitation. In fact,
covering model, maximal covering model, and p-median applications of facility location to emerging markets show
model (Daskin and Dean, 2005). The p-median model aims that several characteristics of these countries are taken into
at minimizing the travelling distance where the farthest consideration at the modelling stage such as travel time, level
consumers can be put at disadvantage, which is usually of education, neighbourhood accessibility, etc. While studies
encountered in rural health facilities for developing nations pertaining to developed countries usually include coverage
(Rahman, 1991). The coverage concept was first introduced distance as the unique decision factor (Mitropoulos et al.,
by Toregas et al. (1971). Using this approach, a customer can 2016). For example, Buor (2003) takes into consideration
be served if a facility is located within a given threshold distance in accessing health facilities in a rural district of
distance or travel time from it (Francis et al., 1992). This Ghana in relation to other factors such as travel time, cost
threshold is called the coverage distance or coverage radius and level of education. Banerji and Fisher (1974) used the p-
(Fallah et al., 2009). Schilling et al. (1993) classify models median and set-covering model to hierarchically locate
which use the concept of covering in two categories: 1) Set facilities for integrated planning in micro-regions of rural
Covering Problem (SCP) where coverage is required and India. In another work, Eaton et al. (1981) applied the
tries to minimize location cost while satisfying a specified maximal covering model to located ambulance bases and
level of coverage. 2) Maximal Covering Location Problem centres from which to recruit rural health workers in
(MCLP) where coverage is optimized and maximizes the Colombia. Another study conducted in India used covering
amount of demand covered within the acceptable service tour models to determine planning mobile health services in
distance by locating a given fixed number of new facilities. the Salcette region of Goa (Hodgson et al., 1998). In Rio,
Facility locations models present differences from one Brazil, Galvão et al. (2002) strategically determined the
sector to another. For example, the private sector’s objective location of healthcare facilities in order to optimize maternal
is generally the minimization of costs or the sum of the and perinatal services. Bapna et al. (2002) also applied
distances to the customers, while the public sector attempts location modelling in India to determine facilities settings of
to apply fairness and equity to all citizens while minimizing unleaded petrol stations using optimization techniques. In a
costs. Then, one of the aims would be the minimization of study in Brazil, public schools’ locations were determined
the distance to the furthest customer. This means that for urban areas using locational analysis and GIS (Pizzolato
distances to other customers may increase so that all et al., 2004). Yasenovskiy and Hodgson (2007) presented
customers benefit from a minimum covering distance (Eiselt applications pertaining to the African country Ghana, for
and Marianov, 2011). The problem of facility location determining the location of healthcare provision bases in
typically arises when dealing with network decentralization rural areas. They proposed a model that incorporates a spatial
decisions because such decisions require determining the choice interaction model attributing attendance and benefits
optimal number and locations of facilities (ex. Mitropoulos to facility size, distance, and neighbourhood accessibility.
et al., 2016). Several governments have attempted to Massam and Malczewski (2016) worked on determining the
decentralize public facilities in order to increase citizens’ location of health centres in a rural region of Zambia using a
service performance levels (Common et al, 2016). decision support system. Another study presented a more
Decentralization has been a prominent organizational reform general case for planning of community health schemes by
in the government activities of emerging markets in recent non-governmental or faith-based organizations in rural areas
years and activities in many sectors, including health, have of developing countries, from both top-down and ground
been decentralized (Bossert et al., 2003; Thunhurst, 2012). level viewpoints (Smith et al., 2009). Rahman and Smith
The effect of a given pattern of decentralization depends on (2000) provide a review of location-allocation models in
the context in which it occurs, including aspects of planning for healthcare facilities, presenting different
geographical configuration, political accountability, applications of mathematical models. They highlight the
technical skill, and cultural norms. As a result, the same differences between developed countries and emerging
Mokrini et al: Modelling Facility Location Problems in Emerging Markets: The Case of the Public Healthcare Sector in Morocco
102 Operations and Supply Chain Management 12(2) pp. 100 - 111 © 2019
markets stating that poor geographical accessibility is the dispersion affect the results of facility location with an
main reason why basic health care does not reach the application to a case study of the Moroccan public
majority of the population in emerging markets. Examples of pharmaceutical supply chain where these characteristics may
their applications pertained to locating hospitals and present an issue.
ambulance stations. In their review, Rahman and Smith
(2000) state that although existing mathematical models of 3. MODEL DESCRIPTION
facility location are too sophisticated for use in emerging
Our solution approach consists of determining the candidate
markets because such setting can present barriers in
locations for facilities first and determining optimal locations
implementation. These models have demonstrated
by solving the developed mathematical model. The first step
usefulness in the facility location decision-making process; of the proposed model approach is to determine candidate
however, they can face political and economic barriers. The
locations for facilities through geographical modelling. For
literature review shows that facility modelling has been the purposes of this study, we selected the set covering model
implemented in different ways in emerging markets, taking
and adapt it to take into consideration characteristics of
specific characteristics that are relevant to each case. The emerging markets. The investigation of literature concerning
main differences in facility location between emerging
the specificities of emerging markets led to choosing road
markets and developed countries are related to geographical infrastructure type and demand dispersion as the main
configuration (difference between urban and rural areas),
characteristics on which the coverage distance would
geographical accessibility, the sector of activity, and nature depend. The aim is to assess how these characteristics affect
of facilities. Although these differences have been
the location modelling in the case of an emerging market.
highlighted in the literature, their impact on the location of Solving the problem allows determining the optimal number
facilities has not been studied. Table 1 presents a
and location of facility locations resulting in a network
summarized comparison between emerging markets and
configuration that satisfies the different requirements of the
developed countries based on characteristics that can affect
problem. Finally, we conduct a sensitivity analysis where
facility location decisions. The characteristics under different scenarios are built to show how the number of
comparison are: geographical configuration, physical
warehouses and the network configuration change when
infrastructure, technological infrastructure, availability of
varying coverage distances taking into consideration road
information, funds and resources utilization, and market
infrastructure type and demand dispersion.
maturity.
Several studies have suggested the negative effects of
emerging market characteristics on supply chain 3.1 Determining the Candidate Locations
management related to the lower quality and availability of The first step of the solution approach is to discretize
infrastructure and resources play an important role in the the location problem to determine candidate locations.
selection of strategic locations (Lorentz et al., 2013). Typically, location problems in continuous spaces are
Emerging markets also present differences compared to modelled using nonlinear optimization problems (where the
developed countries concerning geographical accessibility nonlinearity derives from the Euclidean-distance function),
caused by the differences between rural and urban areas’ it is shown how standard covering problems can be reduced
infrastructures (Khanna and Palepu, 2013). These markets so as to allow a discrete search (Eiselt, 2004). Therefore,
are generally more rural and present a big gap between rural instead of iteratively searching for good solutions in
and urban areas regarding accessibility and resources continuous location problems, like many algorithms or
whereas developed countries are highly urbanized (Lorentz heuristics do, it is suggested to identify a finite set which
et al., 2013). This gap between rural and urban areas also consists of at least one optimal solution. This finite set is
affects demand dispersion of facilities as urban areas are known as the Finite Dominating Set (FDS) (Murray et al.,
more populated. Facility location models present in the 2008). By examining the finite set, instead of the infinite
literature take into consideration different influencing factors space, search effort can be reduced dramatically. Hakimi
such as opening costs, transportation costs, inventories, (1964, 1965) was the first to investigate the FDS for a
demand volumes. In our literature study, we did not come network location problem. He proved that nodes of the
across models that use the road infrastructure type explicitly network comprise an FDS for the p-median problem.
as an input parameter when designing the network. In Church (1984) developed an approach for identifying
addition, the literature on facility location models in the FDS for covering problems involving point-based
emerging markets shows that models have been mainly demand for both Euclidean distance and rectilinear distance.
developed based on problems that exist in developed The FDS is a set of intersection points of the circles with
countries which is not particularly useful when dealing with radius of the service standard centred on the demand
the case of an emerging market failing to take into locations, known as the circle intersection point set (Figure
consideration the differences ins road infrastructure types 1). In order to find the FDS for this model, we use GIS
and demand dispersion. While applications of facility (Geographic Information System) modelling tools to create
location models to emerging markets have taken into buffers around demand centres with distances measured on
consideration characteristics of these countries, no attention the road network. The FDS can be reduced using aggregation
has been given to demonstrating how these differences affect methods. The output of this process is a set of coordinates for
facility location results. This study aims at determining how candidate locations.
differences such as road infrastructure type and demand
Mokrini et al: Modelling Facility Location Problems in Emerging Markets: The Case of the Public Healthcare Sector in Morocco
Operations and Supply Chain Management 12(2) pp. 100 - 111 © 2019 103
Input parameters
J : Set of demand points
I : Set of candidate locations
𝐷𝑖𝑗 : Distance from facility i to demand point j
𝑓𝑖 : Opening cost of facility i
𝑃𝑗 : Category index for demand priority of demand point j
𝑃𝑗 = [1, . . 𝑚] , 𝑚 = 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑎𝑡𝑒𝑔𝑜𝑟𝑖𝑒𝑠
𝑇𝑗 = Category index for road infrastructure type of demand
point j
𝑇𝑗 = [1, . . 𝑛], 𝑛 = 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑎𝑡𝑒𝑔𝑜𝑟𝑖𝑒𝑠
𝑑( 𝑃𝑗 , 𝑇𝑗 ) = Coverage distance function for index 𝑃𝑗 and
index 𝑇𝑗
Figure 1 Circle Intersection Point Set (Mehrez, 2016) 𝑑11 𝑑12 … 𝑑1𝑚
𝑑 𝑑22 … 𝑑2𝑚
𝑑 ( 𝑃𝑗 , 𝑇𝑗 ) = [ 21 ]
3.2 Determining Optimal Locations … … … …
The second step of our approach is the mathematical 𝑑𝑛1 𝑑𝑛2 … 𝑑𝑛𝑚
modelling of the problem with the objective to select the 𝑅𝑗 = 𝑑( 𝑃𝑗 , 𝑇𝑗 ) = Coverage distance required by demand
optimal facility locations from the candidate locations. point j
Following the mainstream in the literature, we opted for a set 1 𝑖𝑓 𝐷𝑖𝑗 ≤ 𝑅𝑗
𝑎𝑖𝑗 = {
covering model (Farahani et al., 2012). The problem seeks 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
to identify a minimal set of facilities and their locations so
Decision variable
that all demand points are covered. The objective is to 1 𝑖𝑓 𝑤𝑒 𝑙𝑜𝑐𝑎𝑡𝑒 𝑡ℎ𝑒 𝑤𝑎𝑟𝑒ℎ𝑜𝑢𝑠𝑒 𝑎𝑡 𝑐𝑎𝑛𝑑𝑖𝑑𝑎𝑡𝑒 𝑙𝑜𝑐𝑎𝑡𝑖𝑜𝑛 𝑖
determine the location of facilities to cover a given set of 𝑋𝑖 = {
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
demand points, such that each demand point is served by a
facility within a maximum service distance Rj. The set Optimization model
covering model minimizes the opening cost of the facilities Minimize ∑𝑖∈𝐼 𝑓𝑖 𝑋𝑖 (1)
that are selected such that all demand nodes are covered. The Subject to:
special feature of this model is that the coverage distance Rj ∑𝑖∈𝐼 𝑎𝑖𝑗 𝑋𝑖 ≥ 1 ∀𝑗 ∈𝐽 (2)
is a function that takes into consideration the emerging
market characteristics. The coverage distance Rj will vary The objective function (1) minimizes the total facilities’
according to the demand points’ dispersion and road cost. Constraint (2) stipulates that each demand node must
infrastructure type. Demand priority is represented by an be covered by at least one of the selected facilities. Finally,
index Pj that can take values 1 to m. For example, demand constraint (3) ensures that 𝑋𝑖 is a binary variable. Kindly
can be categorized to high, medium, or low priority thus note that Rj is a function of the parameters Pj and Tj, which
resulting in 3 varying values for Rj. In the same manner, the denote the ‘demand priority’ and ‘road infrastructure type’
road infrastructure type Tj is an index that can take value 1 to respectively, which affect the optimal facility locations via
n. For example, these categories can be good, medium, and the parameters aij.
bad road infrastructure type. When taking into consideration
Mokrini et al: Modelling Facility Location Problems in Emerging Markets: The Case of the Public Healthcare Sector in Morocco
104 Operations and Supply Chain Management 12(2) pp. 100 - 111 © 2019
The calculation of the opening cost fi is based on the type and a RI ratio that exceeds 40% represented a region
land price and construction costs per squared meter for a with bad road type, values in between were classified as a
typical pharmaceutical warehouse. Existing facilities were region medium quality of road infrastructure. Finally, the
given an opening cost of fi = 0. For other candidate locations, road type index Tj takes three values according to the RI ratio
land price depends on their locations and was retrieved from distribution as shown hereafter:
the real estate price benchmarks provided by the General Tax 1 𝑖𝑓 𝑹𝑰 ≤ 𝟏𝟓%
Directorate of the Ministry of Economy and Finance. 𝑇𝑗 = { 2 𝑖𝑓 𝟏𝟓 % < 𝑹𝑰 ≤ 𝟒𝟎%
Construction costs, on the other hand, are considered equal 3 𝑖𝑓 𝑹𝑰 > 𝟒𝟎%
across the country.
The priority index Pj of demand points was determined 4.2 Results
based on the demand quantity for pharmaceuticals for each The purpose of the model is to determine the optimal
healthcare centre. Demand points that serve a large portion locations of warehouses to cover all healthcare demand
of the population require larger quantities of pharmaceutical points such that each one is served by a facility within a
products and have a high turnover rate. Consequently, maximum coverage distance. The mathematical model is
replenishment of these demand points requires a higher modelled in the General Algebric Modeling Systems
number of trips to deliver the products, which results in (GAMS) software and solved using Cplex solver. The use of
higher transportation costs. Demand data for hospitals and GIS modelling tools in conjunction with the optimization
delegations is available as an aggregate sum of the number program was used for computing data to determine distances
of boxes. For the purposes of this case study, the priority and create buffers, or areas of influence, around centres with
index has been set equal to 3 different values: “high priority”, distances measured on the road network.
“medium priority”, and “low priority”. The cut-off points 4.2.1 Determining candidate locations
were selected dependent on the ranking of the demand The GIS system was used to find candidate locations
volume of each demand point, ensuring a more or less equal according to the coverage distance using intersection points
number of demand points in each category. In order to of the buffers in addition to the current location of hospitals.
determine the priority index of a demand point, the following In order to reduce the number of candidate locations,
steps have been performed: intersection points that fall into a small perimeter were
1- Aggregate demand values for regional/provincial aggregated. Figure 3 shows the buffers (3a), the intersection
hospitals and delegations. This resulted in a total of 81 points and the areas of aggregation (3b) and the final
demand points. candidate locations (3c), respectively.
2- Rank these demand values in descending order 4.2.2 Determining optimal warehouse locations
with (1) being highest demand and (81) the lowest. The set covering model presented in section 3.2 is
3- Determine the cut-off points for the priority index applied to the Moroccan facility location case with sets of 81
Pj, by subdividing the ranking to 3 parts of equal sides: demand points and 49 candidate locations. An integer
𝑃𝑗 =
optimal solution was found by using the branch and cut
1 “High priority” 𝑖𝑓 demand ranking is from (𝟏) 𝐭𝐨 (𝟐𝟕)
algorithm. The distances between candidate locations and
{ 2 “Medium priority” 𝑖𝑓 demand ranking is from (𝟐𝟖) 𝐭𝐨 (𝟓𝟒)
demand points were calculated according to the Vincenty
3 “Low priority” 𝑖𝑓 demand ranking is from (𝟓𝟓) 𝐭𝐨 (𝟖𝟏)
formula. There are several ways to calculate great circle
We are aware that different cut-off points could have distances, including spherical law of cosines, Harversine
been selected but we believe that by doing so, at least, we (Earth as sphere), and Vincenty (Earth as Ellipsoid)
ensure that some priority is given to demand point areas formulas. When dealing with populated areas, the rectilinear
where demand and consumption of pharmaceuticals is high (Manhattan distance) is more accurate than the Euclidean
and requires frequent replenishment. distance. However, when calculating distances over large
The road Infrastructure type of demand points was areas or using latitude and longitude coordinates, distance
determined from the data gathered concerning road networks measurements should be calculated along the curved surface
in Morocco from the Ministry of Equipment and of the earth (great circle distance) (Boscoe, 2013). The
Transportation. This data compares the Moroccan regions Vincenty formula is the most accurate up to date formula that
with values for the number of km of paved roads, non-paved calculates distances between a pair of latitude/longitude
roads, and their percentages compared to total national roads points on the earth’s surface, using an accurate
for every location. The road infrastructure RI ratio has been ellipsoidal model of the earth. The distance between
calculated as follow: candidate location and demand point for our model was
𝑅𝐼 𝑅𝑎𝑡𝑖𝑜 (𝑓𝑜𝑟 𝑒𝑎𝑐ℎ 𝑟𝑒𝑔𝑖𝑜𝑛)
𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑟𝑜𝑎𝑑 𝑘𝑚𝑠 𝑛𝑜𝑛𝑝𝑎𝑣𝑒𝑑 calculated using the Vincenty formula (See Appendix A.)
= In order to evaluate the model, we created 3 sets of
𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑟𝑜𝑎𝑑 𝑘𝑚𝑠 𝑝𝑎𝑣𝑒𝑑
numerical experiments: a base scenario and 2 other sets
When calculating the RI ratio for all regions in the
where demand dispersion and infrastructure type are taken
country, it enabled us to determine two cut-off points (RI =
into consideration respectively. The base scenario represents
15% and RI = 40%) for the RI ratio. Therefore, a region with
an RI ratio less than 15% was considered to have a good road
Mokrini et al: Modelling Facility Location Problems in Emerging Markets: The Case of the Public Healthcare Sector in Morocco
106 Operations and Supply Chain Management 12(2) pp. 100 - 111 © 2019
a b c
Figure 3 Candidate Warehouses Location
the classic set covering with an identical coverage distance warehouses are closer to areas where infrastructure is bad,
for all demand points. Road type and demand dispersion are and further from the areas where road infrastructure is good
not taken into account in this case. By this first set of results such as the far north or the northern east of Morocco. On the
(Table 2), we want to portrait scenarios where 150 km is the other hand, the southern part has a lower infrastructure
average coverage distance. The optimal configuration of the quality compared to the north, which explains the location of
base scenario is illustrated in Figure 4. We can see that 11 warehouses closer to demand points.
warehouses will be opened and distributed evenly around the
areas where hospital centres are located.
10
0
150 200 250 300 350 400
Figure 5 Optimal Warehouse Location for 2nd Scenario Coverage distance (km)
In the third scenario where infrastructure type is taken Figure 7 The Relationship Between the Number of Warehouses
into consideration, 16 warehouses opened. Figure 6 shows and the Coverage Distance
the locations of these warehouses. We can see that the
Mokrini et al: Modelling Facility Location Problems in Emerging Markets: The Case of the Public Healthcare Sector in Morocco
Operations and Supply Chain Management 12(2) pp. 100 - 111 © 2019 107
Figure 8 illustrates the results of scenarios 8, 9, 12, and 200 km (Figure 9 and 10). The results show how the optimal
13, which take into consideration demand dispersion with a number of warehouses changes when the coverage distance
coverage distance varying from 150 km to 350 km and a is varied. The number of warehouses becomes larger when
range of 100 km. The results show that warehouses shift to the range is high. Except when the number of warehouses
urban areas when demand is high. In addition, the number of reaches 3, then the range doesn’t show any difference in the
warehouses varies as the coverage distance is varied as results. The map representation in Figure 11 shows the
shown in Figure 9. results for scenarios 16, 17, 20, and 21 where the range is
Another set of 8 scenarios were constructed by varying 100. The location of warehouses shifts to regions that are less
the road infrastructure type along with the coverage distance urbanized since they present accessibility priority due to a
from 150 to 350 km and the range, taking values of 100 and smaller coverage distance.
1 150 0 0 0 11
2 250 0 0 0 7
3 300 0 0 0 4
4 350 0 0 0 3
5 400 0 0 0 3
6 150 100 1 0 15
7 200 200 1 0 12
8 200 100 1 0 9
9 250 100 1 0 7
10 250 200 1 0 8
11 300 200 1 0 7
12 300 100 1 0 5
13 350 100 1 0 3
14 150 100 0 1 16
15 200 200 0 1 14
16 200 100 0 1 8
17 250 100 0 1 6
18 250 200 0 1 7
19 300 200 0 1 5
20 300 100 0 1 5
21 350 100 0 1 4
22 300 100 1 1 5
23 350 200 1 1 5
24 450 200 1 1 4
25 450 400 1 1 6
Mokrini et al: Modelling Facility Location Problems in Emerging Markets: The Case of the Public Healthcare Sector in Morocco
108 Operations and Supply Chain Management 12(2) pp. 100 - 111 © 2019
10
Tj=1 (GOOD) 300 km 350 km 400 km
5 Infrastructure Type Tj=2 (MEDIUM) 250 km 300 km 350 km
0 Demand Priority Tj=3 (BAD) 200 km 250 km 300 km
200 250 300
Table 5 Coverage Distance Values for Scenario 23
Demand Priority Pj=1 Pj=2 Pj=3
Coverage distance (km) Road Infr. (HIGH) (MEDIUM) (LOW)
Figure 10 Number of Warehouses for Range 200 Km
Tj=1 (GOOD) 350 km 450 km 550 km
Tj=2 (MEDIUM) 250 km 350 km 450 km
Tj=3 (BAD) 150 km 250 km 350 km
Table 6 Coverage distance values for Scenario 24
Demand Priority Pj=1 Pj=2 Pj=3
Road Infr. (HIGH) (MEDIUM) (LOW)
Tj=1 (GOOD) 450 km 550 km 650 km
200 km 250 km Tj=2 (MEDIUM) 350 km 450 km 550 km
Tj=3 (BAD) 250 km 350 km 450 km
Bossert, T. J., D. M. Bowser, and J. K. Amenyah. (2003). “Is Kochetov, Y. (2011). Facility Location: Discrete Models and Local
Decentralization Good for Logistics Systems? Evidence on Search Methods. Combinatorial Optimization. 31, pp. 97-
Essential Medicine Logistics in Ghana and Guatemala.” 134.
Health Policy and Planning 22: pp. 73–82. Kvint, V. (2009). The global emerging market: strategic
Buor, D., 2003. Analysing the primacy of distance in the utilization management and economics. New York: Routledge.
of health services in the Ahafo-Ano South district, Ghana. Laporte, G., Nickel, S., and da Gama, F. S. (2015). Location science
International Journal of Health Planning and Management, (p. 644). Berlin: Springer.
18, pp. 293-311. Lorentz, H., Kittipanya-ngam, P., and Srai, J. S. (2013). Emerging
Carlsson, J. G., and Jia, F. (2014). Continuous facility location with market characteristics and supply network adjustments in
backbone network costs. Transportation Science, 49(3), pp. internationalising food supply chains. International Journal
433-451. of Production Economics, 145(1), pp. 220-232.
Church, R. L. (1984). The planar maximal covering location Massam, B. H., and Malczewski, J. (2016, February). The location
problem. Journal of Regional Science, 24(2), pp. 185-201. of health centers in a rural region using a decision support
Common, R., Flynn, N., and Mellon, E. (2016). Managing public system: a Zambian case study. In Geography Research
services: Competition and decentralization. Elsevier. Forum (Vol. 11, pp. 1-24).
Daskin, M. S. (2011). Network and discrete location: models, Mehrez, A. (2016). Facility Location Problems: Review,
algorithms, and applications. John Wiley and Sons. Description, and Analysis. In Geography Research Forum
Daskin, M. S., and Dean, L. K. (2005). Location of health care (Vol. 8, pp. 113-129).
facilities. In Operations research and health care (pp. 43-76). Melo, M. T., Nickel, S., & Saldanha-Da-Gama, F. (2009). Facility
Springer US. location and supply chain management–A review. European
Drezner, Z., and Hamacher, H. W. (Eds.). (2001). Facility location: journal of operational research, 196(2), pp. 401-412.
applications and theory. Springer Science and Business Ministere de la santé au Maroc. (2014). Termes de references
Media. relatifs à l’étude de faisabilité de l’externalisation du stockage
Eaton, D. J., Church, R. L., Bennett, V. L., Hamon, B. L., and et distribution des produits pharmaceutiques aux
Lopez, L. G. (1981). On deployment of health resources in établissements de santé du Ministère de la Santé. Rapport.
rural Valle Del Cauca Colombia. Studies in the Management Mitropoulos, P., Mitropoulos, I., Giannikos, I., & Sissouras, A.
Sciences, 17, pp. 331-59. (2006). A biobjective model for the locational planning of
Eiselt, H. A., and Marianov, V. (2011). Foundations of location hospitals and health centers. Health Care Management
analysis (Vol. 155). Springer Science and Business Media. Science, 9(2), pp. 171-179.
Fallah, H., Sadigh, A. N., and Aslanzadeh, M. (2009). Covering Mousavi, S. M., Tavakkoli-Moghaddam, R., Heydar, M., and
problem. In Facility Location (pp. 145-176). Physica-Verlag Ebrahimnejad, S. (2013). Multi-criteria decision making for
HD. plant location selection: an integrated Delphi–AHP–
Farahani, R. Z., Asgari, N., Heidari, N., Hosseininia, M., & Goh, PROMETHEE methodology. Arabian Journal for Science
M. (2012). Covering problems in facility location: A review. and Engineering, 38(5), pp. 1255-1268.
Computers & Industrial Engineering, 62(1), pp. 368-407. Mousazadeh, M., Torabi, S. A., Pishvaee, M. S. and Abolhassani,
Francis, R. L., McGinnis, L. F., and White, J. A. (1992). Facility F. (2018), Health service network design: a robust
layout and location: an analytical approach. Pearson College possibilistic approach. International transactions in
Division. operational research, 25: pp. 337–373.
Galvão, R.D., Espejo, L.G.A., Boffey, B., 2002. A hierarchical Murray, A. T., and Tong, D. (2007). Coverage optimization in
model for the location of perinatal facilities in the continuous space facility siting. International Journal of
municipality of Rio de Janeiro. European Journal of Geographical Information Science, 21(7), pp. 757-776.
Operational Research, 138, pp. 495-517. Nickel, S., and da Gama, F. S. (2015). Multi-Period Facility
Haase, K., and Müller, S. (2015). Insights into clients’ choice in Location. In Location Science (pp. 289-310). Springer
preventive health care facility location planning. OR International Publishing.
spectrum, 37(1), pp. 273-291. Pizzolato, N. D., Barcelos, F. B., Lorena, N., and Antonio, L.
Hakimi, S. L. (1965). Optimum distribution of switching centers in (2004). School location methodology in urban areas of
a communication network and some related graph theoretic developing countries. International Transactions in
problems. Operations Research, 13(3), pp. 462-475. Operational Research, 11(6), pp. 667-681.
Han, X. (2014). Healthcare Facility Location and Capacity Rahman, S. (1991). Location-allocation modelling for primary
Configuration under Stochastic Demand (Doctoral health care provision in Bangladesh (Doctoral dissertation,
dissertation, Texas A&M University). University of Exeter).
Hlyal, M., Ait Bassou, A., Soulhi, A., El Alami, J., and El Alami, Rahman, S. U., and Smith, D. K. (2000). Use of location-allocation
N. (2015). Designing a distribution network using a two level models in health service development planning in developing
capacity location allocation problem: Formulation and nations. European Journal of Operational Research, 123(3),
efficient genetic algorithm resolution with an application to a pp. 437-452.
Moroccan retail company. Journal of Theoretical and Rais, A., & Viana, A. (2011). Operations research in healthcare: a
Applied Information Technology, 72(2). survey. International Transactions in Operational
Hodgson, M. J., Laporte, G., and Semet, F. (1998). A Covering Research, 18(1), pp. 1-31.
Tour Model for Planning Mobile Health Care Facilities in Revelle, C. S., Eiselt, H. A., and Daskin, M. S. (2008). A
SuhumDistrict, Ghama. Journal of Regional Science, 38(4), bibliography for some fundamental problem categories in
pp. 621-638. discrete location science. European Journal of Operational
Ibrahim, M. M., and Damasceno, A. (2012). Hypertension in Research, 184(3), pp. 817-848.
developing countries. The Lancet, 380(9841), pp. 611-619. Roberts, M. J. (2011). Pharmaceutical reform: a guide to
Khanna, T., and Palepu, K. (2013). Winning in emerging markets: improving performance and equity. World Bank Publications.
A road map for strategy and execution. Harvard Business Schilling, D. A., Jayaraman, V., and Barkhi, R. (1993). A review of
Press. covering problems in facility location. Location Science, 1,
Khodaparasti, S., Maleki, H. R., Jahedi, S., Bruni, M. E., and 25-55. Pergamon.
Beraldi, P. (2016). Enhancing community based health Schütz, P., Stougie, L., and Tomasgard, A. (2008). Stochastic
programs in Iran: a multi-objective location-allocation model. facility location with general long-run costs and convex short-
Health care management science, pp. 1-15.
Mokrini et al: Modelling Facility Location Problems in Emerging Markets: The Case of the Public Healthcare Sector in Morocco
Operations and Supply Chain Management 12(2) pp. 100 - 111 © 2019 111
run costs. Computers and Operations Research, 35(9), pp. Thunhurst, C. (2012). Public health systems analysis—the transfer
2988-3000. of learning between developed and developing countries.
Smith, H. K., Harper, P. R., Potts, C. N., and Thyle, A. (2009). Health care management science, 15(3), pp. 283-291.
Planning sustainable community health schemes in rural areas Toregas, C., Swain, R., ReVelle, C., and Bergman, L. (1971). The
of developing countries. European Journal of Operational location of emergency service facilities. Operations
Research, 193(3), pp. 768-777. Research, 19(6), pp. 1363-1373.
Snyder, L. V. (2006). Facility location under uncertainty: a review. Yasenovskiy, V., and Hodgson, J. (2007). Hierarchical location-
IIE Transactions, 38(7), pp. 547-564. allocation with spatial choice interaction modeling. Annals of
Şahin, G., and Süral, H. (2007). A review of hierarchical facility the Association of American Geographers, 97(3), pp. 496-
location models. Computers and Operations Research, 34(8), 511.
pp. 2310-2331.
APPENDIX A
Distance (A, B) = ACOS (COS(RADIANS(90 − Lat1)) ∗ COS(RADIANS(90 − Lat2)) + SIN(RADIANS(90 − Lat1)) ∗
SIN(RADIANS(90 − lat2)) ∗ COS(RADIANS(long1 − long2))) ∗ 6371
𝑤𝑖𝑡ℎ Point A = (LAT1, LONG1), Point B = (LAT2, LONG2) and R = mean radius of earth = 6371 km (Boscoe, 2013).
Dr. Asmae El Mokrini is an assistant professor at the business school of the International University of Rabat. Her research
focuses on the design of pharmaceutical supply chains destined for the public sector. She works on different projects in the
field of management science with academicians and practitioners from the industry. Asmae El Mokrini earned her PhD fom
the Mohammadia School of Engineers in Rabat, Morocco and the QUARTZ EA 7393 laboratory in Paris, France. She holds a
Master’s degree in software engineering and management science and a Bachelor's degree in engineering and management
science from Al Akhawayn University in Ifrane.
Youssef Boulaksil has been a faculty member at the College of Business and Economics of the United Arab Emirates
University since September 2013. Prior to that, he was a faculty member at the School of Business Administration at Al
Akhawayn University in Ifrane, Morocco. He holds a MSc in Industrial Engineering and a PhD in Operations Management
and Logistics from the Eindhoven University of Technology, the Netherlands. Youssef also held visiting appointments at the
Nanyang Technological University (Singapore) and the Technical University of Denmark. His research focus is on supply
chain management and logistics, with a focus on emerging markets. He published a number of papers in international journals,
such as: International Journal of Production Economics, IISE Transactions, Interfaces, OR Spectrum, International Journal of
Production Research, and International Journal of Operations and Production Management. Currently, he teaches various
courses in the undergraduate program, as well as in the MBA and the MEM (Master of Engineering Management) programs.
He secured in total over $ 400,000 (USD) of research grants for different research projects in the last few years. For more
information, see: http://faculty.uaeu.ac.ae/y_boulaksil/
Abdelaziz Berrado is Department Chair and Associate Professor of Industrial Engineering in EMI School of Engineering at
Mohamed V University in Rabat. He holds a Ph.D. in Decision Systems and Industrial Engineering from ASU. His research,
teaching and consulting interests are in the areas of Big Data Analytics, Industrial Statistics, Operations and Supply Chain
Modelling, Planning and Control with applications in healthcare, education, manufacturing and other industries. He published
several papers in research journals and conferences with local and international funding. He is a member of INFORMS, IEOM
and IEEE. He was also a Senior Engineer at Intel.